TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 13, No. 3, March 2
015,
pp. 512 ~ 52
0
DOI: 10.115
9
1
/telkomni
ka.
v
13i3.708
9
512
Re
cei
v
ed
No
vem
ber 2
6
, 2014; Re
vi
sed
Jan
uar
y 6, 20
15; Accepted
Jan
uary 20, 2
015
Effect of Shorting Wall on Compact 2×4 MSA Array
Using Artificial Neural Network
Mohammad Anee
sh*
1
, Ja
mshed Asla
m Ansari
2
, Ashish Singh
3
, Km. Kamakshi
4
Micro
w
av
e Ant
enn
a Res
earch
Centre, Dep
a
r
t
ment of Electronics & Comm
unic
a
tion,
Univers
i
t
y
of Allah
aba
d, Alla
h
aba
d – 21
10
02
, UP India
*Corres
p
o
ndi
n
g
author, em
ail
:
aneesh
a
u
1
4
@
gmai
l.com
1
, jaans
ari@r
ediff
mail.com
2
,
ashsi
n09
@red
i
ffmail.com
3
, kamakshi.kum
a
r21@gmail.com
4
A
b
st
r
a
ct
In this p
a
p
e
r,
a co
mp
act 2×
4
microstri
p
a
n
tenn
a (MSA)
array
is pr
es
ented
an
d th
e
effect of
inserti
ng shorti
ng w
a
ll is opti
m
i
z
e
d
w
i
th the
help of
artifici
al ne
ural n
e
tw
ork (ANN). An ANN mo
de
l h
a
s
bee
n
deve
l
o
p
e
d
for
pre
d
ictin
g
the
res
ona
nt
frequ
enci
e
s
w
i
th the
vari
atio
n of th
e
hei
ght
of su
bstrate
a
n
d
shortin
g
w
a
ll
p
o
sitio
n
. F
o
r va
l
i
dati
on
of ANN
outp
u
t, a
pr
ot
otype
of 2×
4 M
SA array
is p
h
ysically
fabric
a
t
ed
and t
e
sted. T
h
is va
lid
atio
n
verifies th
e pr
opos
ed
ante
n
na for w
h
ic
h
simulat
ed
and
ANN res
u
lts
ar
e
appr
oxi
m
ate
l
y similar.
Ke
y
w
ords
: m
i
crostrip antenna array, artificial ne
ural n
e
tw
ork, multib
an
d, shortin
g
w
a
ll
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Over l
a
st
de
cad
e
MSA
s
are
well
a
c
commod
a
ted t
o
wi
rele
ss a
pplication
s
d
ue to it
s
disting
u
ish fe
ature
s
such a
s
lo
w profile
stru
ctur
es, lig
ht weig
ht, low co
st, and e
a
s
y fabri
c
ation
of
different ra
di
ating sh
ape
s [1]. There ha
s bee
n a lot
of rese
arch o
n
MSA in respect to different
sha
p
e
s
. The
sha
pe of
radi
ating pat
ch pl
ays a vita
l rol
e
to enh
an
ce
the ante
nna
cha
r
a
c
teri
stics.
This l
ed to p
r
odu
ce lot
s
of
resea
r
ch work for
different
sha
p
e
s
of MS
As such a
s
, W-sha
ped
pa
tch
[2], E-shap
e
d
patch [3], half E-shap
ed pat
ch [4], S-sh
ape
an
tenna [5], a
nd ga
p coup
led
microstri
p
a
r
ray anten
na [
6
]. MSA arra
ys a
r
e
widely
used i
n
m
a
n
y
pra
c
tical ap
plicatio
ns du
e to
their pl
ana
r
structu
r
e
s
a
n
d
bu
rning
issu
e for re
se
arch am
ong
the
resea
r
chers i
n
p
r
eviou
s
d
a
y
s.
Several inve
stigations
hav
e bee
n do
ne
by re
sea
r
che
r
s
on MSA a
r
rays
su
ch
as,
a si
ngle l
a
ye
r
monop
ulse array [7], cylind
r
ical
array an
tenna [8
], array using
ele
c
tromag
netic
b
and ga
p (EB
G
)
stru
ctures [9-10], and
a
2×2 MSA a
r
ray
with mi
cros
t
r
ip line
feed
to
oth-like-slot
p
a
tche
s [1
1]. In
addition, MS
A array on
an
irregula
r
diel
ectri
c
su
rface
and
Teflon
substrate a
r
e i
n
vestigate
d
[
12-
13]. Furthe
r, desi
gn an
d p
e
rform
a
n
c
e a
nalysi
s
of se
ries fe
ed an
d
corporate fe
ed MSA arra
y
have bee
n carri
ed out [14
]. A spiral array for ultra wideba
nd [15], filtering MSA array [16], a
nd
stacke
d pat
ches fo
r two
dimen
s
ion
a
l
plana
r a
rra
y
topologi
es [1
7] gaine
d m
u
ch
attention
of
resea
r
chers.
All the above
repo
rt
ed
pap
ers
are mainl
y
base
d
on th
e theoretical,
simulatio
n
, a
nd
experim
ental
results. Fu
rth
e
r to increa
se the credi
bili
ty of results it has be
en su
bjecte
d to ANN.
As ANN
provi
des fa
st and
accurate mod
e
lling du
e to
it has gain
ed i
n
tere
st of re
searche
r
s. Th
e
variou
s appli
c
ation
s
of ANN have be
en
utilized in th
e field of MSA arrays
su
ch as, fault finding
[18], frequen
cy modellin
g [19], the effe
ct of mi
cro
s
trip feed line widt
h on arra
y [20], photonic
band g
ap (P
BG) structu
r
e
s
on cylin
dri
c
al MSA [21].
Several othe
r techniq
u
e
s
a
r
e also u
s
ed
for
the de
sig
n
a
nd a
nalysi
s
o
f
MSA array
[22-24]
an
d t
hey p
r
od
uce
fast an
d
accurate
re
sult
s. In
[25], a 64-ele
m
ent array was propo
se
d for Ka-b
and
with rotation feeding te
chni
q
ue.
This
work ha
s bee
n inspired from ab
o
v
e literature
and de
dicate
d to the desi
gn of a
comp
act 2
×
4
MSA array whe
r
ea
s the
effect of inse
rting sho
r
ting
wall on the p
r
opo
se
d ante
nna
array is
opti
m
ized
with
the hel
p of
ANN.
T
here
a
fter p
r
opo
sed ante
nna
is fab
r
icated
and
measured re
sults are
com
p
ared
with ANN
optimizatio
n and sim
u
lat
ed re
sults.
The
re
st of th
e pa
per is o
r
gani
zed
in th
e fo
llo
wing
m
anne
r; sectio
n 2
hold
s
an
overview
of desi
gn a
n
d
simul
a
tion of
antenn
a a
rra
y which al
so i
n
clu
d
e
s
the g
eneration of
simulate
d dat
a
for ANN mod
e
l. Then in
section
3, re
sults
an
d discussion
are
el
aborated fo
r ANN
study a
nd
fabricated p
r
o
posed ante
n
n
a
. Finally in section 4,
the
con
c
lu
sio
n
of the overall st
udy is dra
w
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Effect of Shorting Wall on
Com
pact 2
×
4
MSA Array Using Artifici
al
… (Moh
am
mad Anee
sh
)
513
2. Design an
d Simulation of Antenna
Array
This
sectio
n sho
w
s
the
g
e
o
metry
fo
rma
t
ion
p
r
o
c
e
s
s of propo
se
d antenn
a arra
y. Figure
1(a
)
sho
w
s, 2×4 a
rray of patch ante
n
n
a
s in whi
c
h
each patch i
s
size of
L
×
W
on glass ep
oxy
s
u
bs
tr
a
t
e (
h
=
1.60
mm). All
these patches are
etch
ed
with
L
g
di
stance and
i
n
terconn
ecte
d with
50
Ω
micro
s
t
r
ip lines of
siz
e
s
L
SA
×
W
SA
,
L
SB
×
W
SB
. The total area o
c
cupie
d
by this patch
geomet
ry is
L
FA
×
W
FA
×
h
. Furthe
r, in Fi
gure
1(b
)
, a sho
r
t
ed
wall
is add
ed in radiating p
a
tches
(RP)
one by
one from
RP
1 to RP8. Mu
ltiband resp
o
n
se i
s
a
c
hiev
ed by addi
ng
this shortin
g
wall
as
sh
own
in
Figure 2. All t
he respe
c
tive values
rega
rding
d
e
si
gn specifi
c
ation
s
are su
mmari
zed
in Tabl
e 1.
These a
n
ten
na g
eomet
rie
s
a
r
e
sim
u
la
ted in IE3
D
softwa
r
e [2
6]. The
refle
c
tion
coeffici
ent variation with freque
ncy for shorting
wa
ll p
o
sition
s at RP1 to RP8 is sho
w
n in Fig
u
re
3 an
d o
b
serv
ed that th
ere
is
no
dra
s
tic ch
ang
e o
ccurs in th
e val
ue of
re
son
a
ting fre
que
nci
e
s
only the dip of the reflecti
on co
efficient
is varied
fro
m
-10 dB to -34 dB. Whe
r
eas the effe
ct of
inse
rting sho
r
ting wall on radiating p
a
tches i
s
t
he ma
in cau
s
e for it
s multiple fre
quen
cy ban
d
s
.
Table 1. De
si
gn sp
ecifi
c
ati
ons
Parameter
Value
Sample
L
6 mm
-
W
10 mm
-
L
g
1 mm
-
L
FA
27 mm
-
W
FA
27 mm
-
L
SA
7 mm
-
W
SA
1 mm
-
L
SB
22 mm
-
ε
r
4.7
-
h
1.60 mm
1.57
≤
h
≤
1.6
0
SW
P
RP2
RP1 to RP8
F
e
e
d
in
g
P
o
in
t
L
W
L
g
L
SA
L
SB
L
FA
W
FA
W
SA
(a)
(b)
Figure 1. Simulated de
sig
n
of (a) 2×4 M
SA arra
y (b
) 2×4 MSA array with sho
r
ted wall at RP
2
Figure 2. Refl
ection
coeffici
ent respon
se
of
2×4 MSA a
rray with a
nd
without short
ed wall
Shor
t
i
ng W
a
l
l
F
e
e
d
ing
P
o
in
t
L
W
L
g
L
SA
L
SB
L
FA
W
FA
W
SA
1
2
3
4
5
6
7
8
9
10
-30
-25
-20
-15
-10
-5
0
Fre
q
ue
nc
y
(
G
H
z
)
S
11 (
d
B
)
W
i
th
o
u
t s
h
o
r
ti
n
g
w
a
l
l
W
i
th
s
h
o
r
ti
n
g
w
a
l
l
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 512 – 5
2
0
514
Figure 3. Vari
ation in refle
c
tion coeffici
en
t with frequen
cy for differen
t
shorting
wall
location
2.1. Data
Ge
neratio
n
fro
m
Sim
u
lated Result
s for
AN
N Mod
e
l
This sectio
n
descri
b
e
s
ab
out ge
ne
ratio
n
of
data f
r
o
m
IE3D sim
u
lator fo
r A
N
N mod
e
l’s
training
and
testing. Here
32 anten
na
s are ge
ner
a
t
ed in sim
u
la
tion media
with a variety of
sho
r
ting
wall
po
sition f
r
o
m
RP1
to
RP8
and
he
ight of the
sub
s
trate
(
h
). The
re
son
a
ting
freque
nci
e
s (
f
r
) and
radiatio
n efficien
cie
s
(
R
e
)
of these
32 sim
u
lated
antenn
as a
r
e
record
ed a
n
d
given in Tabl
e 2. These da
ta sample
s a
r
e use
d
for trai
ning an
d testi
ng of ANN m
odel.
Table 2. Simulated data fo
r trainin
g
and
testing of ANN model
SW
P
ε
r
h
L
g
R
e1
R
e2
R
e3
R
e4
R
e5
f
r1
f
r2
f
r3
f
r4
f
r5
Training Data
1
4.7
1.57
1
0.675
0.874
0.714
0.835
0.703
1.31
5.25
6.34
8.26
9.54
2
4.7
1.57
1
0.678
0.734
0.650
0.804
0.745
1.32
5.25
6.42
8.73
9.86
3
4.7
1.57
1
0.574
0.716
0.623
0.810
0.711
1.31
5.32
6.45
8.47
9.70
4
4.7
1.57
1
0.648
0.743
0.641
0.863
0.716
1.30
5.48
6.25
8.46
9.81
5
4.7
1.57
1
0.678
0.754
0.648
0.843
0.730
1.29
5.67
6.43
8.71
9.24
6
4.7
1.57
1
0.699
0.795
0.674
0.817
0.721
1.30
5.65
6.52
8.68
9.41
7
4.7
1.57
1
0.725
0.841
0.645
0.837
0.719
1.35
5.45
6.79
8.64
9.74
8
4.7
1.57
1
0.784
0.803
0.685
0.836
0.701
1.34
5.37
6.63
8.56
9.64
1
4.7
1.58
1
0.642
0.802
0.768
0.812
0.754
1.25
5.61
6.30
8.40
9.90
2
4.7
1.58
1
0.645
0.874
0.688
0.847
0.747
1.29
5.60
6.22
8.39
9.93
3
4.7
1.58
1
0.654
0.897
0.623
0.865
0.753
1.28
5.75
6.25
8.27
9.94
4
4.7
1.58
1
0.754
0.869
0.695
0.811
0.717
1.26
5.74
6.29
8.29
9.93
5
4.7
1.58
1
0.786
0.874
0.684
0.842
0.702
1.22
5.64
6.28
8.37
9.92
6
4.7
1.58
1
0.805
0.841
0.637
0.821
0.715
1.23
5.67
6.27
8.31
9.95
7
4.7
1.58
1
0.745
0.861
0.694
0.823
0.734
1.27
5.81
6.21
8.28
9.91
8
4.7
1.58
1
0.684
0.844
0.678
0.895
0.716
1.29
5.63
6.24
8.38
9.94
1
4.7
1.59
1
0.824
0.874
0.687
0.841
0.736
1.20
5.83
6.23
8.26
9.95
2
4.7
1.59
1
0.812
0.814
0.674
0.845
0.754
1.13
5.85
6.15
8.18
10.1
3
4.7
1.59
1
0.838
0.848
0.669
0.841
0.745
1.18
5.91
6.18
8.13
9.97
4
4.7
1.59
1
0.756
0.804
0.669
0.804
0.700
1.19
5.90
6.22
8.20
9.99
5
4.7
1.59
1
0.778
0.831
0.678
0.845
0.718
1.16
5.86
6.20
8.22
9.98
6
4.7
1.59
1
0.718
0.841
0.645
0.864
0.730
1.17
5.84
6.21
8.13
9.96
7
4.7
1.59
1
0.641
0.867
0.654
0.884
0.732
1.15
5.88
6.17
8.24
9.95
8
4.7
1.59
1
0.653
0.845
0.647
0.888
0.712
1.19
5.87
6.16
8.25
10.0
Testing Data
1
4.7
1.60
1
0.824
0.900
0.724
0.905
0.774
1.23
5.31
6.34
8.62
10.2
2
4.7
1.60
1
0.812
0.905
0.735
0.919
0.780
1.25
5.25
6.42
8.73
9.86
3
4.7
1.60
1
0.838
0.845
0.710
0.906
0.765
1.18
5.38
6.39
8.71
9.60
4
4.7
1.60
1
0.756
0.854
0.707
0.910
0.710
1.11
5.45
6.21
8.66
9.75
5
4.7
1.60
1
0.778
0.896
0.711
0.904
0.768
1.09
5.46
6.32
8.84
9.74
6
4.7
1.60
1
0.718
0.886
0.720
0.898
0.745
1.08
5.42
6.43
8.88
9.63
7
4.7
1.60
1
0.641
0.884
0.704
0.884
0.775
1.14
5.34
6.27
8.79
9.90
8
4.7
1.60
1
0.653
0.890
0.718
0.914
0.740
1.07
5.39
6.48
8.64
9.98
1
2
3
4
5
6
7
8
9
10
-3
5
-3
0
-2
5
-2
0
-1
5
-1
0
-5
0
Fr
e
q
ue
nc
y
(
G
H
z
)
S
11 (
d
B
)
RP
1
RP
2
RP
3
RP
4
RP
5
RP
6
RP
7
RP
8
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Effect of Shorting Wall on
Com
pact 2
×
4
MSA Array Using Artifici
al
… (Moh
am
mad Anee
sh
)
515
2.2. ANN Mo
del Selectio
n for the
An
aly
s
is of Proposed
Anten
n
a
ANN i
s
a co
mputational t
ool (mo
del
) which i
s
in
spire
d
by biologi
cal nervo
us
sy
stem an
d
these n
e
two
r
ks u
s
e
s
sev
e
ral topol
ogi
es. One of
the most po
p
u
lar topol
ogy
is the multilayer
percept
ron
(MLP) ne
ural
netwo
rk
due t
o
its su
pe
rvised and
un
sup
e
rvise
d
lea
r
ni
ng strategie
s
.
So
this is the le
ading rea
s
on
for adapting
MLP-ANN
in this wo
rk.
MLP is a feed forwa
r
d ne
ural
netwo
rk an
d
comp
ri
se
s m
any neu
ro
ns
in input, o
u
tp
ut and
hidd
e
n
layers. Ea
ch layer
plays a
different role
in ANN
mode
l. In this wo
rk, thr
ee laye
re
d MLP-A
NN
model of
stru
cture
9×60
×5
is
sele
cted. It
mean
s this model comp
ri
se
s 9 neuron
s in input layer, 60 neu
ro
ns in hidd
en layer
and 5
neu
ron
s
in the
outp
u
t layer a
s
sh
own i
n
Fi
gu
re
4. The othe
r para
m
eters
are
sele
cted
as,
learni
ng rate (
ƞ
)
= 0.00
1, momentum
coefficient (
µ
)
= 0.75, erro
r
goal (
eg
)
= 0.
0000
1, and
MSE
= 1.21×10
-5
. This MLP model is trai
ned with Lev
enbe
rg Ma
rq
uardt alg
o
rith
m [27] whereas
weig
hts and
biases a
r
e up
dated bet
wee
n
0 and 1. Tw
enty eight data sets fro
m
Table 2 are used
for the
traini
n
g
of A
N
N mo
del, an
d
rem
a
ining
eig
h
t d
a
ta sets
are u
s
ed
for testin
g of A
N
N mo
del.
In this model
simulated d
a
ta pattern, shorting
wall p
o
sition (
SW
P
),
ε
r
,
h
,
L
g
,
R
e1
,
R
e2
,
R
e3
,
R
e4,
and
R
e5
are
use
d
as inp
u
t and their re
spective re
so
n
a
ting frequ
en
cie
s
f
r1
,
f
r2
,
f
r3
,
f
r4
,
and
f
r5
are
con
s
id
ere
d
a
s
o
u
tput fo
r th
is m
odel. T
h
e
n
, du
ring
the t
e
sting
of A
N
N m
odel, m
e
an
squ
a
re
e
r
rors
(MSE) bet
we
en the simul
a
ted and ANN results a
r
e co
mputed a
s
[28]:
2
1
1
[(
)
]
n
iA
N
N
i
i
M
SE
x
f
y
n
(1)
Whe
r
e
x
i
is the final output of ANN mod
e
l
and cal
c
ul
ated as:
21
1
2
21
[
]
(
[
][
]
[
]))
[
]
))
ii
xf
w
f
w
y
b
b
(2)
Whe
r
e
y
i
is the input functi
on matrix and
given as:
12
3
4
5
T
ir
g
e
e
e
e
e
yS
W
P
h
L
R
R
R
R
R
(3)
Whe
r
e
w
1
and
w
2
are the
weig
ht matrices of hidd
en
and outp
u
t layer and given
as:
11
1
1,
1
1
,
2
1,
9
11
1
2,
1
2
,
2
2,
9
1
11
1
60
,
1
60
,
2
60
,
9
..
..
..
.
.
.
..
.
.
.
..
WW
W
WW
W
W
WW
W
(4)
22
1,
1
1
,
5
2
22
30
,
1
30
,
5
..
..
.
.
..
ww
w
ww
(5)
Whe
r
e
b
1
a
nd
b
2
are the bi
as matri
c
e
s
for hidd
en an
d
output layer and given a
s
:
11
1
1
12
6
0
...
T
bb
b
b
(6)
22
2
2
12
5
...
T
bb
b
b
(7)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 512 – 5
2
0
516
SW
P
r
L
g
h
R
e1
R
e2
R
e3
R
e4
R
e5
f
r1
f
r2
f
r3
f
r4
f
r5
w
9,
60
w
1,
1
w
2,
1
w
60
,
5
.
.
b
1
b
2
I
nput
l
a
y
e
r
Hi
dde
n l
a
y
e
r
Out
put
l
a
y
e
r
Figure 4. MLP-ANN m
odel
3. Results a
nd Analy
s
is
This sectio
n
elabo
rate
s th
e expe
rime
ntal resu
lts
of t
he p
r
o
p
o
s
ed
method
which p
r
ovide
s
a sep
a
rate a
n
a
lysis of ANN optimization
and validatio
n study.
3.1. AN
N Re
sults
In the trainin
g
of A
N
N m
odel,
simulat
ed d
a
ta p
a
ttern
s
are
u
s
e
d
a
s
in
put
a
nd o
u
tput
function
s, as
previou
s
ly discu
s
sed. After the prope
r training, AN
N model is
subj
ected to testi
n
g
for IE3D
gen
erated
data f
o
r
sho
r
ted
wall 2×4 MS
A
array. Te
sting data
sets are
sel
e
cte
d
for
sub
s
trate hei
ght
(
h
)
= 1.60
mm and sho
r
ted wall po
sition from
RP
1
to
RP8
as ta
bulated in T
a
ble
2. MLP-ANN model
pre
d
i
c
ted the
ap
p
r
oximatel
y
si
milar re
sults for
re
so
nant freque
nci
e
s and
these results are com
p
a
r
ed with the IE3D sim
u
la
te
d results a
s
given in Tabl
e 3. One of the
tested data
set is physi
call
y fabricated a
nd mea
s
u
r
ed
for validation
of this work and given in the
next se
ction.
The fab
r
icated data
set
h
a
s b
een
hi
ghl
ighted in th
e
Table
2 an
d
Table
3. Figu
re 5
sho
w
s the
compa
r
ison
b
e
twee
n
simul
a
ted a
nd A
N
N o
p
timized
freque
nci
e
s f
o
r
different
test
patterns
.
Table 3.
Co
m
pari
s
on of IE3D, ANN
re
sults
IE3D Simulated (
T
arget
)
ANN output
f
r1
f
r
2
f
r
3
f
r4
f
r
5
f
r1
f
r
2
f
r
3
f
r4
f
r
5
1.23
5.31
6.34
8.62
10.2
1.203
5.312
6.314
8.612
10.202
1.25
5.25
6.42
8.73
9.86
1.253
5.252
6.425
8.734
9.866
1.18
5.38
6.39
8.71
9.60
1.184
5.378
6.389
8.713
9.602
1.11
5.45
6.21
8.66
9.75
1.101
5.445
6.213
8.664
9.745
1.09
5.46
6.32
8.84
9.74
1.092
5.464
6.326
8.854
9.734
1.08
5.42
6.43
8.88
9.63
1.087
5.425
6.437
8.878
9.634
1.14
5.34
6.27
8.79
9.90
1.140
5.334
6.274
8.794
9.907
1.07
5.39
6.48
8.64
9.98
1.073
5.389
6.488
8.634
9.989
Figure 5.
Co
mpari
s
o
n
of simulated an
d ANN fre
que
n
c
ie
s
0
2
4
6
8
10
12
0
2
4
6
8
10
12
T
e
s
t
i
ng P
a
t
t
e
r
ns
Fr
e
que
nc
y
(
G
H
z
)
IE
3
D
(
f
r
1
)
ANN (
f
r
1
)
IE
3
D
(
f
r
2
)
ANN (
f
r
2
)
IE
3
D
(
f
r
3
)
ANN (
f
r
3
)
IE
3
D
(
f
r
4
)
ANN (
f
r
4
)
IE
3
D
(
f
r
5
)
ANN (
f
r
5
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Effect of Shorting Wall on
Com
pact 2
×
4
MSA Array Using Artifici
al
… (Moh
am
mad Anee
sh
)
517
3.2. Validation of ANN Stud
y
For the valid
ation of ANN study a prot
oty
pe of 2×4
micro
s
tri
p
a
n
tenna a
r
ray with a
sho
r
ted
wall
at RP2 i
s
p
h
y
sically fab
r
i
c
ated on
gla
s
s ep
oxy (
ε
r
= 4.7)
su
bstra
t
e as
sh
own
in
Figure 6(a).
An SMA con
necto
r of 0
-
1
8
GHz i
s
u
s
e
d
here for th
e excitation o
f
geometry. T
he
reflectio
n
co
efficient and
radiation p
a
ttern of
fabri
c
ated g
eome
t
ry are mea
s
ured in Agile
nt
N52
30A network a
nalyzer
and an a
n
e
c
h
o
ic chamb
e
r
resp
ectively for provin
g of ANN results.
Figure
7 sh
o
w
s
the com
p
arison betwe
en simula
te
d
and m
e
a
s
ure
d
refle
c
tion
coefficient
(S
11
≤
-10
d
B) for 2
×
4 M
SA array wit
h
a sh
orte
d
wall at RP2.
The figur
e reveal
s
that the
fabricated an
tenna ope
rat
e
s at
multipl
e
freq
uen
cy
band
s
of 1.3
2
, 5.25, 6.
4
2
, 8.73, 9.8
6
GHz,
whi
c
h is
useful for
L
,
C
, a
nd
X
band
ap
plicatio
ns. Si
mulated a
nd
measured re
sults a
r
e in
cl
ose
agre
e
me
nt; only minor div
e
rge
n
ce ta
ke
s pla
c
e
in m
easure
d
valu
e of S
11
due
to the od
dity in
fabrication
a
nd
lo
sses which are not
con
s
id
ere
d
durin
g sim
u
lation. Tabl
e 4 sho
w
s
the
comp
ari
s
o
n
o
f
simulated, ANN, an
d mea
s
ured results
for re
son
a
ting
frequen
cie
s
.
Table 4. Co
m
pari
s
on of si
mulated, ANN, and mea
s
ured
re
sults
Resonating
Freque
nc
y
IE3D Simulated
(GHz
)
ANN
(GHz
)
Experimental
(GHz
)
f
r1
1.25
1.253
1.1
f
r
2
5.25
5.252
5.4
f
r
3
6.42
6.425
6.4
f
r4
8.73
8.734
8.6
f
r
5
9.86
9.866
9.8
(a)
(b)
Figure 6. (a)
Fabri
c
ate
d
2
×
4 MSA Arra
y with RP
2; (b) Experim
en
tal setup for
measurement
of
radiatio
n pattern
Figure 7. Co
mpari
s
o
n
of simulated an
d measur
ed refl
ection
coeffici
ent variation for 2×4 MSA
array with sh
orted wall at RP2
1
2
3
4
5
6
7
8
9
10
-3
0
-2
5
-2
0
-1
5
-1
0
-5
0
Fr
e
que
n
c
y
(
G
H
z
)
S1
1
(
d
B
)
Simu
la
t
e
d
M
eas
u
r
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 512 – 5
2
0
518
Anech
o
ic ch
a
m
ber experi
m
ental
setup for
the
me
asurem
ent of ra
diation patte
rn of 2×4
MSA array
with a sho
r
ted
wall at RP2
is sh
own
in
Figure 6(b) a
nd the di
stan
ce bet
wee
n
the
transmitting a
nd re
ceivin
g
antenn
a is
20
0 cm. In thi
s
work,
E
-plane
(
E
θ
,
ϕ
= 0
˚
) is mea
s
ured in
x-
z
plan
e and
H
-pl
ane (
E
θ
,
ϕ
= 90
˚
) is m
easure
d
in
x-y
plan
e. Com
pari
s
on pl
ots
for simul
a
ted
and
measured ra
diation patterns at ANN o
p
timiz
ed fre
q
uen
cie
s
1.25
3, 5.252, 6.425, 8.734, and
9.866
GHz
shown in Fi
gu
res 8(a)
-(b)-(c)-(d
) and (e
)
re
spe
c
tively
. The
3dB b
e
a
mwidth
is
al
so
estimated a
n
d
summ
ari
z
e
d
in Table 5.
(a)
(b)
(c
)
(d)
(e)
Figure 8. Co
mpari
s
o
n
of simulated an
d m
easured ra
diation pattern at ANN opti
m
ized
freque
nci
e
s
(a) 1.253
GHz (b) 5.25
2 GHz (c)
6.42
5 G
H
z
(d) 8.7
34
GHz (e
) 9.86
6 GHz
-
5
-1
0
-
1
5
-
2
0
-
2
5
d
B
30
210
60
24
0
90
270
12
0
30
0
150
330
180
0
E
-
p
l
an
e
(
s
i
m
u
l
at
ed
)
E
-
p
l
an
e
(
m
eas
u
r
ed
)
H
-
p
l
an
e
(
s
i
m
u
l
at
ed
)
H
-
p
l
an
e
(
m
eas
u
r
ed
)
-
5
-
1
0
-
1
5
-
2
0
-2
5
d
B
30
210
60
240
90
27
0
12
0
300
15
0
330
180
0
E
-
p
l
an
e (
s
i
m
u
l
at
ed
)
E
-
p
l
an
e (
m
eas
u
r
ed
)
H
-
p
l
an
e (
s
i
m
u
l
at
ed
)
H
-
p
l
an
e (
m
eas
u
r
ed
)
-1
0
-
2
0
-3
0
d
B
30
21
0
60
24
0
90
27
0
120
30
0
150
33
0
18
0
0
E
-
p
l
an
e (
s
i
m
u
l
at
ed
)
E
-
p
l
an
e (
m
ea
s
u
r
e
d
)
H
-
p
l
a
n
e (
s
i
m
u
l
at
ed
)
H
-
p
l
a
n
e (
m
ea
s
u
r
e
d
)
-
1
0
-20
-
3
0
d
B
30
210
60
24
0
90
27
0
12
0
30
0
150
330
180
0
E
-
p
l
an
e (
s
i
m
u
l
a
t
e
d
)
E
-
p
l
an
e (
m
ea
s
u
r
e
d
)
H
-
p
l
an
e (
s
i
m
u
l
a
t
ed
)
H
-
p
l
an
e (
m
e
a
s
u
r
e
d
)
-1
0
-
2
0
-30
-4
0
d
B
30
21
0
60
24
0
90
270
120
30
0
150
33
0
18
0
0
E
-
p
l
an
e
(
s
i
m
u
l
at
ed
)
E
-
p
l
an
e
(
m
ea
s
u
r
e
d
)
H
-
p
l
an
e
(
s
i
m
u
l
at
ed
)
H
-
p
l
an
e
(
m
eas
u
r
ed
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Effect of Shorting Wall on
Com
pact 2
×
4
MSA Array Using Artifici
al
… (Moh
am
mad Anee
sh
)
519
Table 5. Co
m
pari
s
on of si
mulated an
d measur
ed be
amwidth
s
at ANN optimi
z
e
d
freque
nci
e
s
ANN optimized
3dB beam
w
i
dth
Angle of Tilt
(for E-
plane)
Radiation Efficie
n
cy
(Simulated)
E
θ
,
ϕ
= 0
˚
(E-pl
a
n
e
)
E
θ
,
ϕ
= 90
˚
(H
-pl
a
ne)
frequenc
y
Simulated
Measured
Simulated
Measured
1.253 GHz
54
˚
52
˚
89
˚
86
˚
-
81.2 %
5.252 GHz
36
˚
35
˚
69
˚
67
˚
-
90.5 %
6.425 GHz
62
˚
59
˚
93
˚
90
˚
30
˚
73.5 %
8.734 GHz
47
˚
49
˚
49
˚
50
˚
33
˚
91.9 %
9.866 GHz
33
˚
28
˚
36
˚
33
˚
27
˚
78.0 %
4. Conclusio
n
In this pap
er, 32, compa
c
t 2×4 MSA
array has
be
en de
sign
ed
in simulatio
n
media
whe
r
ea
s the
effect of in
sertin
g the
shorting
wa
ll i
s
carried
out
with the hel
p of ANN. T
he
develop
ed A
N
N mod
e
l p
r
edicte
d
the
most a
c
cu
rat
e
re
sult
s fo
r re
son
ant fre
quen
cie
s
of
th
e
prop
osed a
n
tenna
array. A comp
act
2
×
4 MSA a
r
ra
y has
been
p
h
ysically fabri
c
ated
on a
gl
ass
epoxy sub
s
trate to validate ANN
study
and re
sult
s
are fou
nd qui
te satisfa
c
tory. The prop
o
s
ed
antenn
a arra
y operate
s
at freque
nci
e
s
1
.
32, 5.25, 6.4
2
, 8.73, and
9
.
86 GHz. Its radiation
patte
rn
has
bee
n me
asu
r
ed
at ANN optimi
z
ed f
r
equ
en
cie
s
. The fab
r
icate
d
anten
na a
r
ray ca
n be
find
good a
pplicat
ions in
L
,
C
, a
nd
X
band.
Referen
ces
[1]
Bahl IJ, Bhartia
P. Microstrip Antenn
as. Artech Hous
e. Ded
h
a
m, Mass. USA. 1980.
[2]
Na
yesta
nak A
A
L. W
-
shape
d
Enha
nce
d
Pat
c
h Anten
na for
W
i
reless C
o
m
m
unic
a
tion.
Wireless P
e
rs.
Co
mmun.
2
0
0
7
; 43:12
57-1
2
6
5
.
[3]
Pand
e
y
VK, Vi
sh
w
a
k
a
rma B
R
. Anal
ys
is of
E-shap
ed P
a
tch Ante
nna.
Mi
crow
ave Opt. T
e
chno
l. Lett
.
200
7; 49:4-7.
[4]
Ansari JA,
Dubey
SK
, Mis
h
ra A. Analy
s
is of Half E-s
hap
ed P
a
tch
for W
i
deb
an
d
Appl
icati
on.
Microwave Opt. Technol. Lett.
2009; 5
1
:157
6
-
158
0.
[5]
Deshm
u
kh AA,
Kumar G. Compact Br
oa
dba
nd S-sha
p
e
d
Microstrip Ante
nnas.
El
ectroni
cs Lett.
2006
;
42:26
0-2
61.
[6]
Meshram MK, Vishvaka
rma
BR. Gap-coup
l
ed Microstrip
Arra
y
Ante
nn
a
for
W
i
de-Ba
n
d
Oper
atio
n.
Internatio
na
l Journ
a
l of Electr
onics
. 20
01; 88
(11):11
61-1
1
7
5
.
[7]
W
ang H, F
a
ng
DG, Che
n
XG
. A Comp
act S
i
ngl
e
La
yer
Mo
nop
ulse
Micro
s
trip Ante
nna
Arra
y
.
IEEE
T
r
ans. on Ante
nnas a
nd Pro
p
ag.
200
6; 54(2)
:503-5
09.
[8]
Svezhe
ntsev AY,
Kr
y
z
h
a
n
o
v
ski
y
V,
Van
d
e
nbosc
h
GAE. C
y
lin
drica
l
Mic
r
ostrip Arra
y A
n
tenn
as
w
i
t
h
Slotted Strip-F
r
amed Patch
e
s.
Progress in El
ectromag
netics
Researc
h
.
201
3; 139:5
39-5
5
8
.
[9]
Gujral M,
Li JL-W, Yuan T
,
Qiu C-W, Band
w
i
dt
h Improv
ement
of Micr
ostrip A
n
ten
n
a
Arra
y us
in
g
Dumm
y
EBG P
a
ttern on F
e
e
d
l
i
ne.
Progr
ess in Electrom
agnetics Research.
2012; 12
7: 79
-92.
[10]
Qiu L, Z
h
a
o
F
,
Xiao
K, Ch
ai
SL, Mao JJ. T
r
ansmit-Rec
e
iv
e Isolati
o
n
Imp
r
oveme
n
t of A
n
tenn
a Arra
ys
b
y
usin
g EBG Structures.
IEEE Antenn
as an
d Wireless Pro
pag. Lett.
201
2
;
11:93-96.
[11]
W
ang H,
Hu
an
g
XB, F
a
n
g
D
G, Han GB. A
Microstrip A
n
te
nna
Arra
y F
o
r
m
ed B
y
M
i
cros
trip Li
ne
F
e
d
T
ooth-Like-Slot
Patches.
IEEE Trans. on Antenn
as an
d Pro
pag.
20
07; 55(
4):121
0-1
214.
[12]
Liu
XF
, Jia
o
Y
C
, Z
hang F
S
,
Che
n
YB.
Des
i
gn of C
onfor
mal Microstri
p
A
n
tenn
a Array
Mounte
d
o
n
a
n
Irregul
ar Die
le
ctric Surface
. Asia Pacific
micro
w
av
e co
nferenc
e proc
eed
ings (AMP
C). 2005; Do
i:
10.11
09/APMC
.
2005.1
6
0
6
5
7
6
.
[13]
Seki T
,
Honma N, Nish
i
ka
w
a
K,
T
s
uneka
w
a
K.
Millim
eter-W
ave Hi
gh-
Efficienc
y Multi
l
a
y
er Par
a
sit
i
c
Microstrip A
n
te
nna
Arra
y o
n
T
e
flon Substra
t
e.
IEEE Trans. on Micr
ow
. Theory
an
d Tec
hni
q.
20
05
;
53(6):2
10
1-21
06.
[14]
Alam MM, Son
c
ho
y MMR, Goni MO.
Des
i
g
n
an
d Perfor
mance A
nalys
is
of Microstrip A
rray Anten
n
a
.
Progress i
n
Ele
c
tromagn
etics Rese
arch S
y
m
posi
u
m Procee
din
g
s
.
200
9; Mosco
w
,
Russ
ia,
1837-
18
42.
[15]
Al
w
a
n EA, Sertel K, Volak
i
s JL. A Simple Eq
ui
val
ent Circ
u
it
Model for Ultr
a
W
i
deb
an
d C
oup
led Arr
a
ys.
IEEE Antennas
and Wirel
e
ss Propa
g. Lett.
2012; 11:
117-
12
0.
[16]
Lin
CK, Ch
un
g
SJ. A F
ilterin
g
Microstrip A
n
t
enn
a Arra
y.
IE
EE T
r
ans. on
Microw
. T
heor
y and
T
e
ch
niq.
201
1; 59(1
1
):2
856-
286
3.
[17]
Vazqu
e
z C, H
o
topa
n G, Ho
e
y
e
SV, F
e
rna
d
e
z
M, Herra
n
LF
, Heras F
L
.
Microstrip A
n
te
nna
Des
i
g
n
Based o
n
Stacked Patches fo
r Reconfig
ura
b
l
e
T
w
o Dim
ens
i
ona
l Plan
ar Arra
y
T
opol
ogi
es.
Progress i
n
Electro
m
a
gneti
cs Researc
h
. 2009; 97:
95-1
0
4
.
[18]
Patnaik A, C
h
oud
hur
y B, Pr
adh
an P, Mish
ra RR, Christ
o
dou
lou
C. An
ANN App
lic
ati
on for F
ault
F
i
ndi
ng in A
n
te
nna Arra
ys.
IEEE Transac. on Ant
ennas and Propag.
200
7; 55(3):77
5
-7
77.
[19]
Patnaik
A, Anagnos
tou D,
Christodoulou CG
, Ly
k
e
JC. M
o
deling Frequenc
y
Reco
nfigurable Antenna
Arra
y
us
in
g Ne
ural N
e
t
w
orks.
Microw
ave Opt. T
e
chnol. Lett.
2005; 4
4
(4):35
1-35
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 3, March 2
015 : 512 – 5
2
0
520
[20]
Dun
dar O, Uz
er D, Gultek
in
SS, Ba
y
r
ak M.
Effects of mic
r
ostrip fee
d
l
i
n
e
w
i
dth o
n
1×
4 recta
ngu
la
r
micr
ostrip ant
e
nna arr
a
y elect
r
ical par
a
m
eter
s and esti
mati
o
n
w
i
th artificial neur
al netw
o
rk
s
. Progress
in Electrom
agn
etic Rese
arch
S
y
mp
osi
u
m Pr
ocee
din
g
s. 201
2; KL, Mala
ysi
a
, 1341-
13
45.
[21]
W
ang
X, Z
h
a
n
g
M, W
ang
SJ
. Practibil
i
t
y
A
nal
ys
is
a
nd A
p
plicati
o
n
of PB
G Structures o
n
C
y
l
i
ndr
ical
Conform
a
l Micr
ostrip Anten
na
and Arra
y.
Pro
g
ress in El
ectromag
netics Re
search
. 20
11;
115:4
95-
507.
[22]
Lee K
C
. F
r
eq
uenc
y-
Doma
in
Anal
ys
es of
Nonl
in
earl
y
Loa
de
d Anten
na Arra
ys us
i
ng Simu
late
d
Anne
ali
ng Al
go
rithms.
Progres
s in Electro
m
a
gnetics R
e
sear
ch
. 2005; 5
3
:2
71-2
81.
[23]
He QQ, Wang BZ. Design of Microstrip Arra
y
A
n
tenna by
Us
ing Element Pattern T
e
chniqu
e
Combi
n
i
ng
w
i
t
h
T
a
y
l
or S
y
nth
e
sis Metho
d
.
Progress i
n
Elec
troma
g
n
e
tics R
e
searc
h
. 200
8; 80:63-7
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.